Test Bed
Processor |
AMD Opteron
144 CACJE 0605BPBW |
CPU
Cooler |
Thermalright
SI-120 Thermalright XP-90 |
Thermal
Paste |
Arctic Ceramique |
Motherboard |
DFI nF4 Ultra-D |
Memory |
2 x 512MB
OCZ Value VX |
Graphics
Card |
Inno3D 6800GS |
Power Supply |
Enermax Noisetaker
535W |
Casing |
Cooler Master
Centurion 530 |
Operating
System |
Windows XP
Profesional SP2 |
Other Components |
Cooler Master
Aerogate II |
Testing Methodology
Three methods were used to obtain temperature measurements.
The first method was to use Motherboard Monitor 5 configured to read the temperatures on a DFI motherboard. The second method involved a temperature probe connected to a Cooler Master Aerogate II panel. It was placed at the CPU socket.
Whilst the temperature readings will obviously not reflect the actual CPU temperature because of the distance between the probe and the core, it should give a fairly consistent temperature reading which will help to show the difference in cooling capacity. A second probe was placed in front of the case's intake fan to measure the ambient temperature.
Team ARP uses delta temperatures in our result analysis. The CPU temperature was taken and subtracted against the ambient temperature to get the difference or delta. This eliminates fluctuations in the ambient temperature. The delta measurement therefore provides a more consistent and accurate result by eliminating variables due to fluctuations of the ambient temperature.
The third method of measurement involves testing the overclockability of the CPU against the various cooling configurations tested. This method was first expounded at SgOverclockers, where Sin22 and EazTerence used this method to test heatsinks and watercooling products. Their tests showed that lower temp readings do not necessarily translate into better overclockability.
Additional Info
Throughout the temperature tests, the CPU was overclocked to 2990 MHz (332 MHz x 9) with the voltage set to 1.5V in the BIOS. Socket 939 Opterons have been observed to run fairly warm and the overclocking should increase its heat output quite significantly.
Idle temperatures were measured by letting the system boot into Windows XP and leaving it alone for 15-20 minutes. For load temperatures, Prime95's Small FFT test was used to heat up the CPU. PWM temperatures were also recorded through MBM5 since they are directly affected by the CPU cooler's airflow.
For the overclockability test, the maximum frequency at which the CPU can complete the Super Pi 8M test (a calculation of Pi to 8 million digits) was set as the bar to meet. This test is fairly consistent in testing CPU clock limits, and it only takes about 5 minutes. Do note, however, that the completion of Super Pi 8M may not guarantee a fully stable system. It's just serves as a measuring tool for our tests.
For this test, it was important to ensure that all factors, other than the heatsinks, remained as constant as possible. The heatsinks were tested at the same ambient temperatures, and steps were taken to ensure that other factors such as the memory modules did not limit overclockability.
The case has two Delta 120mm fans (WFB1212M rated for 72.39cfm) at the front and the rear, with the rear fan running at 7V. There is an additional 80mm side intake fan rated for 32.4cfm. The power supply has a 92mm intake fan and an 80mm exhaust fan. The heatsinks were tested in a closed case to simulate a typical usage scenario.
We pitted the SI-120 against the XP-90. These were the fans used for the respective heatsinks :
SI-120 |
XP-90 |
|
Fan |
Sanyo Denki
|
Delta |
Model |
109R1212H1011 |
AFB0912HH |
Dimensions |
120 x 38mm |
92 x 25mm |
Current Draw |
0.52 A |
0.25 A |
Power |
6.24 W |
3 W |
Speed |
2600 rpm ±
10% |
3200 rpm ±
10% |
Air
Flow |
102.5 cfm |
57.92 cfm |
Noise
Level |
39 dBA |
38 dBA |
The Sanyo Denki fan was tested at 3 settings (full rpm, 2000rpm and 1400rpm) while the Delta fan was tested at 2 settings (full rpm, 1800rpm) to simulate fans with different rpm/airflow levels.
Arctic Ceramique was the thermal paste used throughout the tests and efforts were made to ensure that its application was as consistent as possible.
Now, let's take a look at the test results!